It is your requirement that something must succeed in a properly conducted double blind test.
VFB or CFB is just a name. What important is a finished amplifier. Why not just give a name of your preferred CFB amp, and your preferred VFB amp. As simple as that. Or just say if you cannot perceive any different between any 2 amplifiers. That's why I'm sure you cannot succeed in the mineral water double blind test. But can't you see, I still have a believe that you can pass the "one month test"?
What I think I've accepted so far (in generic terms):
cfb faster rise time
cfb slightly less phase shift compared to a typical ltp (in the >10Mhz range)
vfb lower dc offset (typical)
noise was mentioned at some point but I dont think there was any pursuit of it.
Aside from the input stage the rest of the circuitry can be identical.
So it may be that the common usage of "cfb" and "vfb" dont really tell the entire story.
It appears more like the cfb symmetrical input may just be more compatible or easy to implement with other design approaches which result in lower gain at the expense of higher bandwidth.
Hope this helps
-Antonio
cfb faster rise time
cfb slightly less phase shift compared to a typical ltp (in the >10Mhz range)
vfb lower dc offset (typical)
noise was mentioned at some point but I dont think there was any pursuit of it.
Aside from the input stage the rest of the circuitry can be identical.
So it may be that the common usage of "cfb" and "vfb" dont really tell the entire story.
It appears more like the cfb symmetrical input may just be more compatible or easy to implement with other design approaches which result in lower gain at the expense of higher bandwidth.
Hope this helps
-Antonio
Or may be it is not just ears that we use? May also be skin, etc
Yes, I have been thinking for years about what makes us enjoy the music. What physical parameters responsible for that.
First, it is "sonic". I found correlation with current and voltage swing (ADD: and "damping factor" of course). For mosfet amps (notice the high gate capacitance), it is clear to me the effect of current drive.
Second, it is "fatigue". There is no good about good sound if it disturbs you. This one is a mystery. We know a little about IMD/TIM etc. From experience with working and voicing passive crossovers, I suspect it is related with phase, which has relationship with feedback. Non NFB amps sound good without FB, but they sound bad because of lack of drive. How about sufficient drive without introducing FB?
I don't know theory as good as you all do, but I found a "red line" to believe in CFB, at least from examples of CFB amps. If you can make a better VFB amp, just make it, I will then try to find out why the VFB amp is superior than CFB amp.
Last edited:
Or may be it is not just ears that we use? May also be skin, etc
Yes, I have been thinking for years about what makes us enjoy the music. What physical parameters responsible for that.
First, it is "sonic". I found correlation with current and voltage swing. For mosfet amps (notice the high gate capacitance), it is clear to me the effect of current drive.
Second, it is "fatigue". There is no good about good sound if it disturbs you. This one is a mystery. We know a little about IMD/TIM etc. From experience with working and voicing passive crossovers, I suspect it is related with phase, which has relationship with feedback. Non NFB amps sound good without FB, but they sound bad because of lack of drive. How about sufficient drive without introducing FB?
I don't know theory as good as you all do, but I found a "red line" to believe in CFB, at least from examples of CFB amps. If you can make a better VFB amp, just make it, I will then try to find out why the VFB amp is superior than CFB amp.
actually the ears are just an input port, it all happens in the brains, where the electrical signals from the ears are passed on.....
people are unique individuals, each gifted with a brain.....
in my own experience, i found that CFB amps packs a punch in the bass department but not so good in the mids and highs where the emitter followers do better....i based my comments from building Daniel Meyer's Tigersaurus amps and the Leach double barreled amps....YMMV...
What double blind testing does is to eliminate as far as possible what people think they hear and isolate what they actually do hear.
It has been shown time and again what people see and expect is as much to do with what they say they hear, as what they actually do hear.
Being a scientist I want to tease the whole thing out and isolate what we can actually hear, because you simply can't trust your ears to tell you what you hear because what you hear is a subjective thing bound up with all sorts of other things.
As has been pointed out me being me and you being you mean that we listen to the same thing but hear different things, especially if we know about the source and how it looks and so do the people who are doing the test.
If you don't have all of this in the way then you are left pretty much with what goes into the ear and is processed by the brain into the sensation of sound.
And when you do this audio types who make the sort of claims they do can't actually hear what they say they hear, and as I have pointed out before usually claim that the test must be invalid, I am a scientist however and one disadvantage of that is that I must accept the hypotheses that best fits the data, despite how ever much it contradicts my prejudice.
rcw
It has been shown time and again what people see and expect is as much to do with what they say they hear, as what they actually do hear.
Being a scientist I want to tease the whole thing out and isolate what we can actually hear, because you simply can't trust your ears to tell you what you hear because what you hear is a subjective thing bound up with all sorts of other things.
As has been pointed out me being me and you being you mean that we listen to the same thing but hear different things, especially if we know about the source and how it looks and so do the people who are doing the test.
If you don't have all of this in the way then you are left pretty much with what goes into the ear and is processed by the brain into the sensation of sound.
And when you do this audio types who make the sort of claims they do can't actually hear what they say they hear, and as I have pointed out before usually claim that the test must be invalid, I am a scientist however and one disadvantage of that is that I must accept the hypotheses that best fits the data, despite how ever much it contradicts my prejudice.
rcw
i refrain from making my own comments on things that i build.....i let others do it, i keep quiet......
if they like what they heard, i am happy for them, if not......i don't really care, but i do listen to their comments and then try to see what tweaks i can do.....
i DIY stuff mainly to please my ears....not to impress others..... and certainly not to make claims based on a single event....
if they like what they heard, i am happy for them, if not......i don't really care, but i do listen to their comments and then try to see what tweaks i can do.....
i DIY stuff mainly to please my ears....not to impress others.....
and certainly not to make claims based on a single event....
I don't think you can tell by listening the difference between VFB and CFB.
A friend and I, working on two totally different designs with VFB have gotten the epic soundfield size common to Singleton input and similar types, but his NTP, and mine LTP. In comparison the CFB, the VFB takes more time and more precision on compensations. It also seems that the VFB has more limited gain. I know I've definitely had to use lower gain with VFB and it is definitely memorable because all of the cases have been inconvenient. In comparison, the CFB offered faster gratification.
Of the variances that happen during fine tuning, the CFB is more likely to do poor frequency response (which is easier to fix) and the VFB is more likely to do shortchanged soundfield size (which takes longer to fix and is harder to measure). SO, the labor with VFB involves shots in the dark until finally hitting the target, and the labor with the CFB is faster and more measurable. Of course both can do hi-fi, but how hard do you want to work for it? In several ways, VFB means more labor. But, it is doable.
I think that it is a case of choosing the tools you're most comfortable with.
The VFB people are going to be making some rather nice preamplifiers, and the CFB people are going to be making some rather nice buffers. Perhaps the labor, at end result, isn't much different after all.
A friend and I, working on two totally different designs with VFB have gotten the epic soundfield size common to Singleton input and similar types, but his NTP, and mine LTP. In comparison the CFB, the VFB takes more time and more precision on compensations. It also seems that the VFB has more limited gain. I know I've definitely had to use lower gain with VFB and it is definitely memorable because all of the cases have been inconvenient. In comparison, the CFB offered faster gratification.
Of the variances that happen during fine tuning, the CFB is more likely to do poor frequency response (which is easier to fix) and the VFB is more likely to do shortchanged soundfield size (which takes longer to fix and is harder to measure). SO, the labor with VFB involves shots in the dark until finally hitting the target, and the labor with the CFB is faster and more measurable. Of course both can do hi-fi, but how hard do you want to work for it? In several ways, VFB means more labor. But, it is doable.
I think that it is a case of choosing the tools you're most comfortable with.
The VFB people are going to be making some rather nice preamplifiers, and the CFB people are going to be making some rather nice buffers. Perhaps the labor, at end result, isn't much different after all.
Last edited:
???
It is exactly the contrary.
See here the difference, on the same amp between the two configurations: www.esperado.fr - Le crescendo revisité
It is easy to understand why, as current feedback remove one pole in the loop (the input inverting stage) , it increase the bandwidth.
It is easy to understand why, because Current .Feedback is directly removed from the original signal, resultant signal to be amplified has lower level -> lower distortion. And why the inverting stage of a voltage feedback, adding his own distortion to the feedback signal; add total distortion.
And, even on a good amp schematic, you can really and definitively hear the difference. Both in preamps and power amps.
Curent feedback has advantages in near all the domains: Open loop bandwidth (Slew rate), HF phase shift, H.D. and IM, so... transparency. And only one negative side effect: reduced PSSR that you can cure with accurate schematic design.
Thanks for the explanation.
In my above comments; however, I was thinking about setting the amplifier closed loop gain to a proportion of the open loop gain--this is a Lower gain setting for VFB if you want transparency. The lower closed loop gain setting may be inconvenient. If so, use CFB.
In my above comments; however, I was thinking about setting the amplifier closed loop gain to a proportion of the open loop gain--this is a Lower gain setting for VFB if you want transparency. The lower closed loop gain setting may be inconvenient. If so, use CFB.
You are right on this. It is easy to understand that TIM is in direct proportion of the open loop bandwidth. On fast transients, if the amp is not able to follow the signal's speed, it introduce ugly TIM distortion. That is the reason why Feedback has so bad (unjustified) reputation.Thanks for the explanation.
In my above comments; however, I was thinking about setting the amplifier closed loop gain to a proportion of the open loop gain--this is a Lower gain setting for VFB if you want transparency. The lower closed loop gain setting may be inconvenient. If so, use CFB.
With current feedback, you can easily reach >3Mhz bandwidth, > 500V/µs of slewrate. and, so, 0° of phase error at 20 Khz. Here, more feedback you set, less distortion in the audio range and better damping factor.
The good practice, once you have tuned your amp (CFB or VFB) to maximize bandwith, is to add a low pass filter in the input, to ensure the amp will never have to deal with faster edges (transients) than he can deal with.
Comparing CFB vs VFB, the difference in sound impression is better transparency with CFB, better dynamic, in a strange way, more solid basses (that is because big transients are mainly in basses and Kick drums), more fluid treble and ease with dynamic.
Increased bandwidth lead to stronger and more separated basses and less aggressive treble impression.
Christophe
If the designs are similarly optimized for speed (lets say essentially identical except for the input stage), then is it really the extra phase shift from the inverting side of a vfb that limits the bandwidth (and are there limits such as the driving source impedance)?
Is the error really different between the two, isn't it still just the non-linear voltage required to change the current into the current to voltage stage?
Does the class-B slew rate performance of the cfb ever come into play for audio signals?
Just trying to understand,
Thanks
-Antonio
If the designs are similarly optimized for speed (lets say essentially identical except for the input stage), then is it really the extra phase shift from the inverting side of a vfb that limits the bandwidth (and are there limits such as the driving source impedance)?
Is the error really different between the two, isn't it still just the non-linear voltage required to change the current into the current to voltage stage?
Does the class-B slew rate performance of the cfb ever come into play for audio signals?
Just trying to understand,
Thanks
-Antonio
As you say, in the link i published, the two version are identical, except the feedback path.
You can simulate, (here simulation tools help a lot to understand) to see the signals differences in the feedback line.
About current vs voltages, it is just a matter of (bad) naming things. Because of the Ohm law, you can look at what happens both from voltage or current point of view. It is the same thing.
In fact, in the so called "current feedback" version, the feedback kind of modulate the gain of the first stage. And, because the bandwidth of the first stage is proportional to his gain factor, you can understand why bandwidth is increased. Not to forget than the low impedance of the feedback line help to minimize influence of parasitic capacitances at high frequencies.
One of the very interesting features of CFB is that the bandwidth remain nearly constant, whatever gain of the amp or feedback factor. For Current Feedback amps, the loop gain is set by the feedback impedance allowing an independent setting for the signal gain. The feedback impedance becomes the frequency response compensation.
You can simulate, (here simulation tools help a lot to understand) to see the signals differences in the feedback line.
About current vs voltages, it is just a matter of (bad) naming things. Because of the Ohm law, you can look at what happens both from voltage or current point of view. It is the same thing.
In fact, in the so called "current feedback" version, the feedback kind of modulate the gain of the first stage. And, because the bandwidth of the first stage is proportional to his gain factor, you can understand why bandwidth is increased. Not to forget than the low impedance of the feedback line help to minimize influence of parasitic capacitances at high frequencies.
One of the very interesting features of CFB is that the bandwidth remain nearly constant, whatever gain of the amp or feedback factor. For Current Feedback amps, the loop gain is set by the feedback impedance allowing an independent setting for the signal gain. The feedback impedance becomes the frequency response compensation.
Analog Design, Texas Instrumets, etc.. have published many papers to explain the theory of both VFB and CFB. Just Google for Voltage feedback vs Current feedback.
http://www.ti.com/lit/an/snoa376a/snoa376a.pdf
http://www.ti.com/lit/an/slva051/slva051.pdf
http://www.ieee.li/pdf/viewgraphs/current_feedback_vs_voltage_feedback_amplifiers.pdf
etc...
http://www.ti.com/lit/an/snoa376a/snoa376a.pdf
http://www.ti.com/lit/an/slva051/slva051.pdf
http://www.ieee.li/pdf/viewgraphs/current_feedback_vs_voltage_feedback_amplifiers.pdf
etc...
It is not necessarily generally true: you can use the additional loop gain provided by the VFB to extend the closed loop bandwidth.
This may not always be possible or desirable, but it is a theoretical option.
The level has nothing to do in this, and anyway the level is not dependent on the input stage topology but on the voltage gain of the stages comprised between the input stage output and the speaker terminals
Generally, the opposite will be true. If we look at the typical CFB example provided by Catalin:
the main source of non-linearity (in the amplifier itself, leaving the IP buffer aside) will be the non-linearity of Q2's B-E.
When it is orphaned, as in this example, it uncompensated and outside the FB loop.
But when the amplifier has been converted to VFB:
a similar non-linearity is inserted into the FB path and provides a first order compensation
I'd like to see the results of fully randomized fully blind A B tests...
For a general purpose building block like an IC, this is a selling point.
For a discrete amplifier you design yourself, the usefulness of such a feature is more debatable, unless you want to locate the volume control in the feedback network
Please, in that matter, believe the old sound engineer i was.
And i know the material i'm listening too, because it can be my own mixages.
I do not believe so much in blind tests neither. For numerous reasons too long to detail here.
I partially agree, with some remarks.
One is you can set bandwidth to his maximum with CFB and keep-it there.
Second you can play with local gain of each stage VS loop global gain to optimize distortion.
Third, you can tune damping factor to your needs without affecting bandwidth..
One example where VFB is a no no is in the stereo bus of a mixing desk. More you plug tracks in it, more you increase the gain of the mixing stage, more you reduce its bandwidth, and get a dull sound. One example where CFB change your life.
I have no time, for now, to fight again your other controversial argues, it would be too long and boring. But some of your you asserts are interesting, like "similar non-linearity" compensation. I use that kind of perversity to increase bandwidth, using current mirror in the first stage because efficiency decrease after FT, and so more current for the same level of HF, more level at the output: local auto compensation. One of my answers would be to talk slew rate instead of closed loop bandwidth.
The main one is, if you prefer VFB for some obscure reason, chose your poison. My own religion is made since more than 30 years, correlated by numerous experiences, objective or subjectives, and many measurements .
Those who are interested with CFB can read this very interesting thread: http://www.diyaudio.com/forums/solid-state/193923-simple-symetrical-amplifier.html and read how DIYers talk about the sound of their CFB amps in a very similar manner.
PS: in your schematic, Feedback resistance do not have a proper value for CFB. Too high impedance. And the feedback is not applied to the first stage; witch limit the bandwitch by itself. And here, you apply CFB to a x1 voltage gain stage...
Last edited:
I find it incomprehensible that you are an engineer, and yet do not "believe" in double blind tests.
If a test you do is not double blind then it is subject to systematic error such that it has an in built bias that biases the outcome to the one you wanted in the first place.
All that an increased sample size does is to reinforce this bias.
Statisticians know this well and marketers use single blind tests for precisely this reason, to give "scientific", respectability to what is marketing hype.
You gloss over your objections to double blind testing, I am curious as to why.
rcw
If a test you do is not double blind then it is subject to systematic error such that it has an in built bias that biases the outcome to the one you wanted in the first place.
All that an increased sample size does is to reinforce this bias.
Statisticians know this well and marketers use single blind tests for precisely this reason, to give "scientific", respectability to what is marketing hype.
You gloss over your objections to double blind testing, I am curious as to why.
rcw
THD and DF need no no "tuning", simply minimization for one and maximization for the other. DF can easily be tuned (degraded) by adding a series resistor.
No perversity in that, simple application of sound engineering practices
I believe in the right tool for the right job.
I am under the impression that CFB has a (mostly unjustified) aura, whereas VFB is accused of all evils of the earth.
When you look at the subject from a purely objective point of view, the picture is far from that kind of manicheism.
.
The source schematic isn't mine, it is an example provided by a keen CFB supporter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.